KR101610610B1 - Encoder, decoder and method - Google Patents

Abstract

There is provided an encoder 10 for encoding input data D1 comprising a sequence of numerical values to generate corresponding encoded output data D2 or D3, ) Comprising a data processing device for applying a form of difference and / or sum encoding to generate one or more corresponding encoded sequences, wherein the one or more corresponding encoded sequences are encoded by the encoded output data D2 Or D3), the maximum value is wrapped around and / or the minimum value is wrapped, and the encoder 10 generates a series of < RTI ID = 0.0 > May be operated to use a first default prediction value for a prediction value and the encoded data is generated using an input value, a prediction value and an encoding operator .

Description

[0001] ENCODER, DECODER AND METHOD [0002]

The present disclosure relates to an encoder, e.g. an encoder, which can be operated to use a direct ODelta operator. The present disclosure also relates to a method of encoding data, for example, a method of encoding data by using a direct ODelta operator. The present disclosure also relates to a decoder that decodes the encoded data, such as a decoder that can be operated to apply an inverse ODelta operator. Additionally, this disclosure relates to a method of decoding encoded data, for example, a method of decoding encoded data by applying an inverse ODelta operator. Additionally, the present disclosure relates to a software product recorded on an non-transient (non-volatile) machine-readable data storage medium, wherein the software product can be executed on computer hardware for implementing the methods described above.

Claude E. Shannon proposed a mathematical theory of communication that provided the basis for modern communication systems. In addition, various modern encoding methods have been developed within the knowledge of the mathematical theory described above. A list of information sources providing an overview of modern technical knowledge is provided in Table 1.

The known technical literature Early literature                      Detail
P1 "Variable Length Code", Wikipedia (accessed November 28, 2012)
URL: http://en.wikipedia.org/wiki/Variable-length_code
P2 "Run-Length Encoding", Wikipedia (accessed November 28, 2012)
URL: http://en.wikipedia.org/wiki/Run-length_encoding
P3 "Hoffman coding", Wikipedia (accessed November 28, 2012)
URL: http://en.wikipedia.org/wiki/Huffman_coding
P4 "Arithmetic coding", Wikipedia (accessed November 28, 2012)
URL: http://en.wikipedia.org/wiki/Arithmetic_coding

P5 "A Mathematical Theory of Communication", Shannon, Claude E. (1948) (accessed November 28, 2012)
URL: http: //cm: bell-labs.com/cm/ms/what/shannonday/shannon1948.pdf
P6 "Delta Encoding", Wikipedia (accessed November 28, 2012)
URL: http://en.wikipedia.org/wiki/Delta_coding
P7 Shannon's source coding theory; Wikipedia (accessed November 28, 2012)
URL: http://en.wikipedia.org/wiki/Source_coding_theorem
P8 "Entropy" - Wikipedia (accessed November 28, 2012)
URL: http://en.wikipedia.org/wiki/Entropy

In European Patent EP 1 2376 974 B1 ("Method and Apparatus for Packet Header Compression", Alcatel Lucent (FR)), a method of transmitting a data packet is described, wherein the data packet contains a field CF . The compressed value represents a value developed between two consecutive data packets and includes a predetermined interval (hereinafter " interpolation interval "). The method comprises the following steps:

(i) appending an additional bit to the compressed value if the distance between the compressed value and the predetermined wraparound boundary is less than a predetermined threshold, wherein the additional bit is a relative position of the compressed value Wraparound boundary, the method comprising the steps of:

(ii) at the receiver, calculating a first interpretation interval according to all received bits of the compressed value,

(iii) signaling a wraparound if a value greater than one value of the first analysis interval coincides with the received compressed value,

(iv) at the time of signaling of the wrap-around, calculating a second interpretation interval in accordance with all received bits of the compressed value, excluding additional bits,

(v) clarifying the decompressed value in the second analysis interval using an additional bit.

The definitions of Shannon entropy are provided in documents P7 and P8 listed in Table 1. There are a plurality of different compression methods that can be operated to compress the entropy present in a given data, and these methods are sometimes used to change entropy, e.g., to obtain a larger lossless compression ratio for a given data; This entropy modification method may be used, for example, for run-length encoding (RLE) as described in document P2 of Table 1, variable length coding (VLC) as described in document P1 of Table 1, Hoffman Coding, delta coding as described in document P6 of Table 1, and arithmetic coding as described in document P4 of Table 1, i.e., range coding. These methods are advantageously used to compress data representing alphabetic characters, numbers, bytes and words. However, these methods are not well-suited for compressing given data at the bit level, and for this reason are not very suitable for compressing such given data which may be changed to bit-by-bit.

For example, the delta encoding described in document P6 of Table 1 may be operated to generate delta values that may be positive or negative values from the positive original data values. There is also an implementation of a known delta encoder used for 8-bit, 16-bit or 32-bit wrap around, based on the size of the data elements used. However, modern delta encoders are lacking which are optimized for regimes other than 8-bit, 16-bit or 32-bit values. In particular, known delta encoders generally use the coding original bit values as " 0 ", " 0 ", " &Quot; and " 1 ", respectively.

All kinds of data consume storage space, and communication system transmission capacity is needed when such data is used from one location to another. As the amount of data increases as a result of multimedia developments such as, for example, three-dimensional video content, more storage space and transmission capacity are required to process the data, and more energy is required as the amount of data increases . Globally, the amount of data communicated is incrementally increased over time; For example, the Internet contains huge amounts of data, some of which are stored in multiple copies. Also, there are currently available methods for compressing the entropy (E) associated with the data, e.g., for use in reducing the size of the data. There are also available methods for changing entropy, for example, for delta encoding and run-length encoding (RLE), but improved methods are needed to provide even greater data compression that is currently feasible.

For example, in a bit-by-bit encoding, a previous or subsequent encoding method in which all values of data elements in the original data are not used and / or used in conjunction with a delta encoding method, There is also a need for an optimal utilization of known delta encoding methods that allows for faster and more efficient encoding of the original data, in the case of requiring higher bit formats than dynamic.

This disclosure attempts to provide a delta encoder, or " direct ODelta encoder, " which is an improved type, which encodes different values of data as well as individual bits, i.e., to be.

The present disclosure also attempts to provide an improved method of delta encoding data, i. E., A method of direct ODelta encoding data.

The present disclosure also attempts to provide an improved decoder, i.e., an inverse ODelta decoder, for decoding encoded data, such as ODelta encoded data.

Additionally, the present disclosure attempts to provide an improved method of decoding encoded data such as ODelta encoded data, i. E., An inverse ODelta decoding method of data.

According to a first aspect, there is provided an encoder as claimed in appended claim 1, wherein an encoder for encoding input data (D1) comprising a sequence of numerical values for generating corresponding encoded output data (D2 or D3) Wherein the encoder comprises a data processing device for applying a type of difference and / or sum encoding to the input data (D1) to generate one or more corresponding encoded sequences, wherein the one or more corresponding encodings The resulting sequence is wrapped and / or wrapped around a maximum value to produce encoded output data D2 or D3 and the encoder 10 generates encoded output data D2 or D3 And the encoded output data may be operable to use a first default prediction value for a set of prediction values used for the input value, e.g., It is generated using the value and encoding operator.

Advantageously, the wraparound is used to avoid generating negative differences or too large a value when implementing an encoder. The term " numerical value " will be interpreted to include a separate bit (binary) as well as a group of bits (higher order than binary).

The present invention is based on the finding that the combination of difference and / or sum encoding and wrap around can provide useful entropy changes of the input data D1 to produce encoded data D2, It is advantageous in that improved data compression is achieved within a given entropy encoder when generating the generated data D3.

Optionally, for the encoder, the data processing device may calculate one or more offsets, minimum and / or maximum values for applying to the one or more corresponding encoded sequences used to generate the encoded output data (D2 or D3) To analyze the input data D1 and / or one or more corresponding encoded sequences. Optionally, for the encoder, one or more offset values have a value of " 0 ".

Optionally, the encoder is operated to process a numerical value comprising one or more 1-bit values, and the encoder is operated to encode the input data D1 in a bit-by-bit manner.

Optionally, in the encoder, one or more corresponding encoded sequences represent changes in sequential values of the input data D1.

Optionally, the encoder operates to use the wrap around value (wrapValue), and the wrap around value is the maximum value (highValue) - the minimum value (lowValue) + 1. This wrap around value is used to avoid generating these values when generating the encoded output data D2 or D3, which is a minimum value (lowValue), often smaller than the negative value, or higher than the maximum value (highValue) , And is beneficially used.

Optionally, the encoder can be operated to subdivide the input data D1 into a plurality of data sections that are individually encoded. Optionally, the encoder may also be operable to selectively apply the encoding to the data section only when the data compression can be achieved in the encoded output data (D2 or D3).

Optionally, in the encoder, the first default prediction value is " 0 ". Optionally, the first prediction value may also be a lowValue, a highValue, a highValue + a lowValue + 1/2, or the like.

Optionally, the encoder may be operable to apply additional encoding to generate the encoded output data D2 or D3, where the additional encoding may be a run-length encoding (RLE), a split RLE , An entropy modifier (EM), variable length coding (VLC), Hoffman coding, arithmetic coding, and range coding.

Optionally, the encoder is capable of encoding efficiently using run length encoding (RLE), split RLE, entropy modifier (EM), Huffman coding, variable length encoding (VLE), range coding and / And may be operated to subdivide the input data D1 into a plurality of sections according to the run-length of similar bits.

Optionally, for an encoder, the processing device is implemented using computing hardware operative to execute one or more software products recorded on a non-transitory machine-readable data storage medium.

According to a second aspect, there is provided a method of using an encoder to encode input data (D1) comprising a sequence of numerical values to generate encoded corresponding output data (D2 or D3)

(a) using an encoder's data processing device to apply the type of difference and / or sum encoding to the input data (D1) to generate one or more corresponding encoded sequences;

(b) using the data processing device to wrap the surroundings and / or wrap the surroundings in the one or more corresponding encoded sequences to produce the encoded output data (D2 or D3).
The method includes using a first default prediction value for a series of prediction values used to generate encoded output data (D2 or D3), wherein the encoded data is encoded using an input value, a prediction value and an encoding operator .

Advantageously, wraparound is used to avoid generating a negative difference, or too large a value, when implementing an encoding method. The term " numerical value " will be interpreted to include individual bits (binary) as well as groups of bits (higher order than binary).

Optionally, the method further comprises the step of generating input data D1 (D2 or D3) to calculate one or more offsets, minimum and / or maximum values for applying to one or more corresponding encoded sequences for use in generating the encoded output data D2 or D3 ) And / or a data processing device that analyzes one or more corresponding encoded sequences. Optionally, the method also includes using a " 0 " value for one or more offset values.

Optionally, the method comprises processing a numerical value comprising one or more 1-bit values and encoding the input data D1 in a bit-by-bit manner. Optionally, the method also includes using a " 0 " value for one or more offset values.

Optionally, when implementing the method, the one or more corresponding encoded sequences represent changes in sequential values of the input data D1.

Optionally, the method comprises operating an encoder to use a wrap around value, wherein the wrap around value is a highValue - lowValue + 1. This wraparound is advantageous in order to avoid generating these values when generating the encoded output data D2 or D3, which is a minimum value (lowValue), often less than the negative value or higher than the maximum value (highValue) Lt; / RTI >

Optionally, the method includes subdividing the input data D1 into a plurality of data sections that are individually encoded. Optionally, the method further comprises selectively applying the encoding to a section of data only when the data compression can be achieved in the encoded output data (D2 or D3).

Optionally, in the method, the default first predicted value is " 0 ". Optionally, the first prediction value may also be a lowValue, a highValue, a highValue + a lowValue + 1/2, or the like.

Optionally, the method comprises applying additional encoding to generate encoded output data D2, wherein the additional encoding includes at least one of run-length encoding (RLE), split RLE, entropy modifier (EM) Variable length coding (VLC), Hoffman coding, arithmetic coding, and range coding.

Optionally, the method further comprises the steps of: using a run-length encoding (RLE), split RLE, entropy modifier (EM), Hoffman coding, variable length encoding (VLE), range coding and / And subdividing the input data D1 into a plurality of sections according to the run length of the input data D1.

Optionally, the method also includes implementing the processing device using computing hardware operable to execute one or more software products recorded on a non-transitory machine-readable data storage medium.

According to a third aspect, there is provided a decoder for decoding encoded data (D2, D3 or D4) to produce decoded output data (D5), said decoder comprising one of the encoded data (D2, D3 or D4) Wherein the data processing device is operable to apply a type of difference and / or sum decoding to one or more corresponding encoded sequences of the one or more portions, wherein the one or more encoded The sequence is wrapped around and / or wrapped around a minimum value to produce decoded output data D5, and the data processing apparatus is configured to estimate a default value of a first estimate in a series of data to be decoded Can be operated.

Optionally, the decoder may be operable to decode the encoded data (D2, D3, or D4) comprising one or more 1-bit values, and the decoder may decode the bit-by-bit encoded data D2, D3 Or < RTI ID = 0.0 > D4. ≪ / RTI >

Optionally, the decoder is arranged such that the data processing apparatus can be operated to receive one or more values for applying to one or more encoded sequences for use in generating decoded output data D5. Optionally, the received values are a maximum value, a minimum value and / or an offset value. Optionally, the decoder can also be operated on a function for one or more offset values having a value of " 0 ".

Optionally, the decoder may be operated on a function for one or more corresponding encoded sequences indicating a change in sequential values encoded with the encoded data (D2, D3 or D4).

Optionally, the decoder may be operated to use a wrap around value (wrapValue), wherein the wrap around value is equal to the maximum value (highValue) - minimum value (lowValue) + 1. This wraparound is used to avoid generating these values when generating the decoded output data D5, which is a minimum value (lowValue), often less than the negative value or higher than the maximum value (highValue).

Optionally, the decoder is operable to cause the data processing device to perform processing via at least one of run length encoding (RLE), split RLE, entropy modifier (EM), variable length coding (VLC), Hoffman coding, arithmetic coding, Lt; RTI ID = 0.0 > data. ≪ / RTI >

Optionally, the default prediction value also has a value of " 0 ". However, the prediction value may also be a lowValue, a highValue, a highValue + a lowValue + 1/2, etc., as described above.

Optionally, for a decoder, the processing device is implemented using computing hardware operable to execute one or more software products recorded on a non-transitory machine-readable data storage medium.

In a fourth aspect, there is provided a method of using a decoder to decode encoded data (D2, D3 or D4) to produce corresponding decoded output data (D5), the method comprising the steps of: , D3, or D4), the data processing apparatus comprising: means for generating a difference and / or sum decoding type on one or more corresponding encoded sequences of one or more portions And wherein the one or more encoded sequences are wrapped around and / or wrapped at a minimum value to produce decoded output data (D5), the data processing apparatus comprising: Lt; RTI ID = 0.0 > 1 < / RTI >

Optionally, the method comprises decoding the encoded data (D2, D3 or D4) comprising one or more 1-bit values and decoding the encoded data (D2, D3 or D4) in a bit- Lt; / RTI > decoder. Optionally, in the method, the one or more offset values have a value of " 0 ".

Optionally, the method may comprise input data D1 for applying to one or more corresponding encoded sequences for use in generating decoded output data D5 and / or one or more corresponding encoded sequences D2, D3 Or < RTI ID = 0.0 > D4. ≪ / RTI >

Optionally, in the method, the one or more corresponding encoded sequences represent a change in sequential values encoded with the encoded data (D2, D3 or D4).

Optionally, the method includes using a data processing device wraparound the maximum value or wrap around the minimum value to avoid generating a negative difference or too large a value when generating the decoded output data D5 do.

Optionally, in the method, the data processing device is adapted to perform at least one of run length encoding (RLE), split RLE, entropy modifier (EM), variable length coding (VLC), Hoffman coding, arithmetic coding, Lt; RTI ID = 0.0 > and / or < / RTI >

Optionally, in the method, the default prediction value has a value of " 0 ".

Optionally, in the method, the processing device is implemented using computing hardware operable to execute one or more software products recorded on a non-transitory machine-readable data storage medium.

According to a fifth aspect, there is provided a method of using a decoder to decode encoded data (D2, D3 or D4) to produce corresponding decoded output data (D5)

(a) comprises at least one encoded sequence of encoded data (D2, D3 or D4) representing a change in a sequential value of the transformed data, and wraparound the maximum value or wrap around the minimum value (D2, D3, or D4) to decode one or more portions of the encoded data (D2, D3, or D4), using the data processing device to process the encoded data

(b) a data processing device that generates corresponding processed data and transforms one or more portions using at least one pre-offset value and / or post-offset value to produce decoded output data D5 Lt; / RTI >

With reference to the above-described embodiment, for the offset value, optionally, at the decoder, the data processing device receives a range of values in the encoded data (D2, D3 or D4), from which at least one pre- And / or to derive a post-offset value. The post-offset value may be used to generate the transformed data prior to use of the inverse operator, e.g., the direct inverse ODelta operator, and the pre-offset value is used to transform the data after the inverse operator. The at least one offset value is optionally combined with the processed data to produce decoded output data D5.

Offsets are optionally supported in the decoder. Offset is used when this is also used in the encoder. The wraparound, inverse operator (difference vs. difference) and inverse predictor (input value versus output value) in the parameter (wrapValue) using the range (lowValue to highValue) are important components of the decoder.

Optionally, in the decoder, the data processing device is adapted to perform at least one of run length encoding (RLE), split RLE, entropy modifier (EM), variable length coding (VLC), Hoffman coding, arithmetic coding, Lt; RTI ID = 0.0 > and / or < / RTI > This processing is executed to generate the data D4 from the data D3.

For a decoder, the data processing device may be operable to use wraparound within the parameter wrapValue using a range (lowValue to highValue) when implemented in the form of reverse decoding, e.g., inverse ODelta decoding.

According to a sixth aspect, at least one encoder according to the first aspect for encoding input data (D1) to generate corresponding encoded data (D2 or D3), and generating corresponding decoded data (D5) There is provided a codec comprising at least one decoder according to the third aspect for decoding encoded data (D2, D3 or D4).

According to a seventh aspect, there is provided a software product recorded on an non-transient (non-transient) machine-readable data storage medium, the software product being executable on computing hardware to perform a data encoding method according to the second aspect .

According to an eighth aspect, there is provided a software product recorded on an non-transient (non-transient) machine-readable data storage medium, the software product being executable on computing hardware to execute a data decoding method according to the fourth aspect .

It will be appreciated that the features of the invention may be combined in various combinations without departing from the scope of the invention as defined by the appended claims.

Embodiments of the present disclosure will now be described by way of example only with reference to the following drawings.
1 is a diagram illustrating a codec including an encoder and a decoder implemented for functionality according to the present disclosure;
Figure 2 is a diagram illustrating the steps of a data encoding method as implemented in the encoder of Figure 1;
3 is a diagram illustrating the steps of a data decoding method as implemented in the decoder of FIG.
In the accompanying drawings, an underlined number is used to indicate an item with an underlined number or an item with an underlined number adjacent to the underlined number. An underscore number is associated with an item identified by a line connecting an underscored number to the item. If a number is underscored and accompanies an associated arrow, the underscore number is used to identify the generic entry pointed to by the arrow.

In describing the embodiments of the present disclosure, the following abbreviations and definitions are used, as provided in Table 2 below.

Abbreviations and definitions Abbreviation Explanation ADC Analog-to-digital converter Codec The encoder and corresponding decoder for digital data DAC Digital / Analog Converter
DB A database in a random access memory (RAM) or a read-only memory (ROM)
DC Represents the minimum spatial frequency component of the image, corresponding to the DC component of a given image, i.e., the means of the image, i.e., the average luminance RLE Run-length encoding ROI Areas of interest ROM Read only memory VLC Variable length code

In overview, referring to FIG. 1, the present disclosure relates to an encoder 10 and its associated method of operation, and advantageously, the encoder 10 is implemented as a direct ODelta encoder. The present disclosure also relates to a corresponding decoder 20, and advantageously, the decoder 20 is implemented as an inverse ODelta decoder. Embodiments of the present disclosure advantageously use a range-optimized version of other data as well as a direct ODelta operator, which is a bit-optimized version of the known Delta encoding method described above. The ODelta encoding may be performed using variable-length data words such as 8/16/32/64 bits and / or variable-length encoding of 8/16/32/64 bit data elements whose original values are represented in the range of 1 to 64 bits Or dedicated digital hardware, and the corresponding encoded value is generated from 1 to 64 bits. Of course, the encoder 10 and the decoder 20 know that in some cases some kind of numerical value is included in the data D1, for example in the original data, its definition or transmission is not further described here. It is simply assumed that the numerical ranges MIN and MAX are known, and data D1 can be used.

The known delta coding method increases the range of values from the original (MIN to MAX) to the result (MIN-MAX to MAX-MIN). This means that this method also produces a negative value when the original data contains only a positive value. The ODelta operator according to the present disclosure never generates a value that is not in the range of the corresponding original value, so this operator does not increase the data range used, and accordingly at runtime, for example, at the time of entropy reduction and associated data compression . ≪ / RTI > For example, the known delta encoding method is based on the fact that the data values generated by this delta encoding method can be substantially represented using bits ranging from -31 to +31, i.e., 6 (i.e., sine bits + 5 bits) Bit data, that is, a range of values from 0 to 31, and, conversely, direct ODelta generated values are generated from the stream of 5-bit data described above, 0 to 31. < / RTI > Also, while known methods of delta encoding are not allowed to be implemented recursively, direct or vice versa, the ODelta operator according to this disclosure is allowed to be recursively implemented, but this operator continues to use the values of the range used . The range of values does not have to be bit-exact, for example values of 0 to 31 are defined by 5 bits, and the ODelta operator can use ranges of values of any range, for example from 0 to 25 On the other hand, it still works properly.

In principle, the ODelta method described here can always function directly based on the existing data range, which is given by way of example below. The ODelta method may also be improved by conveying information indicating a minimum occurring numerical value (&quot; lowValue &quot;) in the data and a maximum occurring numerical value (&quot; highValue &quot; It should be noted that lowValue> = MIN and highValue <= MAX and these values are optional.

Two examples of the direct and inverse ODelta operators according to the present disclosure are described below. A first example of direct and inverse ODelta operators may be implemented in electronic hardware and / or computing hardware that can be operated to execute one or more software products, for example, written on non-transient (non-volatile) machine-readable data storage media It is efficient and relatively simple.

When implementing the direct or inverse ODelta operator according to the present disclosure, advantageously the entirety of the original sequence of data values is positive, and the lowest value is zero. Optionally, some offset values, i.e., pre-offset values or post-offset values, can be used to shift the data values such that they all have positive values and the lowest value is &quot; 0 &quot;. The ODelta operator according to the present disclosure can be used for all types of data in a direct manner; This can generally provide data compression, i.e., reduce the communicated data rate, since when the offset value is added to all values or subtracted from all values, the range of data values is a fractional bit As shown in FIG. For example, prior to the application of the direct or inverse ODelta operator, the original data values are in the range of -11 to +18, this range may be modified to a range of 0 to 29 using an offset value of +11, Lt; RTI ID = 0.0 &gt; 5-bit &lt; / RTI &gt; When these pre-offset or post-offset values are not used, the original data values require at least 6 bits to describe them, and sometimes, in practice, a full 8-bit sign byte is used for convenience.

Similar optimization of the data range is also possible when using the generalized direct or inverse ODelta operator. Accordingly, if a direct or inverse ODelta operator, or some other method, generates a data value that can be provided with an offset value that is less than the full-range value, then optimization of that range may be implemented in some phase in the ODelta encoding method . When a negative or positive in the offset value, sign, is used, it must also be transmitted from the encoder 10 to the decoder 20, as will be described later with reference to Figs. 1, 2 and 3.

The direct ODelta operator can be implemented in a one-bit manner and can encode the original data D1 in a bit-by-bit manner, for example, and in this one-bit scheme, as will be described in more detail below, 1 and 3 generate a value &quot; 0 &quot; when there is no change in the bit value of the original data D1 in Fig. 1 and generate a value &quot; 1 &quot; when a change in the bit value of the original data D1 occurs do. The prediction of the first bit in the original data is optionally a value of &quot; 0 &quot;, so that the value of the first bit in the original data D1 is maintained. It is alternatively optionally possible to use the first bit of the predicted value in the original data, but such a selection does not provide any coding advantage, since the prediction is optionally performed on the 1- By default, the selection does not need to be transmitted when it is always assumed to be a value of &quot; 0 &quot;, that is, a predetermined value &quot; 0 &quot; is used by the encoder 10 and the decoder 20, Thereby improving data compression. &Lt; RTI ID = 0.0 &gt;

An example of a direct ODelta encoding according to the present disclosure will now be described. An exemplary original bit sequence, 27 bits including 17 &quot; 1 &quot; and 20 &quot; 0 &quot;, is provided in Equation 1 (Eq.

식 1

Equation 1

Its entropy (E) can be calculated from Equation 2 (Eq.2).

식 2

Equation 2

The number of bits required to code the entropy (E) in Equation 2 (Eq. 2), that is, Min_bit

. Can be calculated from Shannon's source coding theory, as described in D7 and D8 above, and is provided in Equation 3 (Eq. 3).

식 3

Equation 3

The original sequence of bits includes 37 bits in which there are 13 &quot; 1 &quot; s and 24 &quot; 0 &quot; s when processed by the Direct ODelta operator as described above, i.e. Method 1 and Method 3, .

식 4

Equation 4

Its entropy (E) can be calculated from Equation 5 (Eq. 5).

식 5

Equation 5

This can be expressed by the minimum number of bits, Min_bits, according to Equation 6 (Eq. 6).

식 6

Equation 6

The sequence of bits in Equation 4 (Eq. 4) may advantageously be used to achieve data compression using, for example, at least one of runle Tm encoding (RLE), Huffman coding, arithmetic encoding, range encoding, entropy modifier encoding or SRLE encoding And is further encoded.

The ODelta operator can be used in place of the original data, for example as in Equation 4 (Eq. 4), for example when RLE or SRLE is applied to the data in which the associated entropy coding method is applied, Bit direct ODelta operators, i.e., Method 1 and Method 3, reduce the amount of bits required to represent the original data D1, which is equal to the number of bits in the original sequence of bits in Equation 1 (Eq. 1) &Quot; 1 &quot; when there are many changes, and the operator generates &quot; 0 &quot; when there is a long sequence of mutually-similar bits in the original sequence of bits in Equation 1 (Eq.

The inverse version of the ODelta operator, that is, the inverse of method 1 and method 3, modifies the bit value from a value of "0" to a value of "1" appropriately or from a value of "1" to a value of "0" 0 &quot; in the encoded stream of data D2 when there is a &quot; 1 &quot; value in the data D2, i.e., in the stream of encoded data, Do not change the bit value when present. When this ODelta operation is performed on the direct ODelta-computed bit stream of data D2, the original stream of data D1 is generated as decoded data D5, but as described above, VLC or Hoffman coding and Such additional coding is beneficially used and this coding also needs to be considered because data D3 is generated from data D2 using the forwarding operation of the entropy encoder and data D4 is generated from the entropy decoder And is generated from the data D3 using the inverse operation.

Advantageously, the original stream of data D1 is subdivided into two or more sections before the encoding is applied. These subdivisions provide an opportunity for a greater number of optimizations used when encoding the original stream of data D1. For example, such a subdivision is beneficial because the alterable sequence in data D1 is directly ODelta encoding, that is, it produces more &quot; 1 &quot; when using Method 1 and Method 3, whereas a flat, non-alterable sequence, quot; flat &quot; sequence preferably generates more &quot; 0 &quot; for subsequent VRL or Hoffman encodings, Section can be reduced for the entire bit stream constituting the data D1.

An example of a direct ODelta encoding according to the present invention will now be described when a plurality of sections that are individually encoded from each other are used. The first section including the sequence of the original single bits includes 16 bits in total, that is, 7 &quot; 1 &quot; and 9 &quot; 0 &quot; in Equation 7 (Eq.

식 7

Equation 7

Where H (x) = 4.7621 and B = 15.82; "H" represents entropy, and "B" represents Max_bit. When the original 7-bit sequence (Eq. 7) of the original bit is processed by the direct ODelta operator, the corresponding sequence of converted bits is provided as in Eq. 8 (Eq. 8).

식 8

Equation 8

Here, H (X) = 4.3158 and B = 14.34.

The second section containing the sequence of original single bits is included in Equation 9 (Eq. 9) as follows.

식 9

Equation 9

Here, H (X) = 6.3113 and B = 20.97. The sequence of the original bit of Equation 9 (Eq. 9), when processed by the direct ODelta operator, provides the corresponding sequence of transformed bits as in Equation 10 (Eq. 10).

식 10

Equation 10

Where H (X) = 1.7460 and B = 5.80. In these examples, H (X) denotes entropy E, and B denotes the minimum number of bits required for coding, as described above.

Optimal compression from Eqs. 7 (Eq. 7) and Eq. 10 (Eq. 10) in this example is achieved when both sections are processed separately with a direct ODelta operator (ie 14.34 bits + 5.80 bits = 20.14 bits total For encoding); This compression requires less than the 36.82 bits required by the original, i.e., the number of original ODelta-calculated bits requiring 34.60 bits or the number of original bits required after division (= 15.82 bits + 20.97 bits = 36.79 bits). Advantageously, dividing the original stream of bits in the data D1 into sections, i.e., entropy (E) of the original data D1 and modified data (D), as contained in the data D2 by piece- Lt; RTI ID = 0.0 &gt; (H) &lt; / RTI &gt;

The data compression is performed in a coarse manner by simply dividing the portions of the data D1 into new sections to be encoded when there are a plurality of long run sections available in the data D1, Where the bit values change rapidly according to the sequence. Optionally, some sections of data D1 are encoded without using the direct ODelta operator, for example if there is a long run of mutually-similar bits with relatively few individual difference bits; In this case, the direct ODelta operator does not provide a significant advantage for data compression purposes.

Dividing the data D1 into smaller sections has the disadvantage of creating additional overhead which contributes to making the data into encoded data D2. Such overhead includes, for example, information indicating the amount of data bits or data bytes associated with each new section. However, it is always found that at least some amount of overhead data value needs to be transmitted, so that there is only one extra overhead data value when the given data is divided into two sections of data.

Entropy encoding is advantageously implemented after a direct ODelta operator, e.g., VLC, Huffman coding, arithmetic coding, range coding, RLE, SRLE, EM and similar coding, to achieve an encoded bitstream that can be decoded thereafter. It is easier and computationally more efficient to perform the optimization calculation based on the calculated entropy (E) and the minimum bit estimate value when compared to the actual data encoding. This order of execution enables considerable speed optimization and often achieves optimal data compression results in the encoded data D2. Alternatively, it is realized to perform entropy optimization in such a way that the original bit, alphabet, number, byte and word data in data D1 are first coded in some other way to produce an entropy-optimized bitstream And then the direct ODelta operator is used to modify the entropy-optimized bitstream to provide the corresponding encoded data, i.e. data D2. Further, this ODelta-calculated data can be continuously encoded from data D2 in a different encoding method to generate data D3.

The generalized direct ODelta operator uses a value or bit number of a bit, which is a parameter that describes a range of values used in the data D1, that is, the bits necessary to represent the values. In addition, the ODelta operator is used in a way that can use positive and negative offset values, that is, positive and negative &quot; pedestal &quot; values. For example, if 7 bits are provided in the data D1, that is, it has a value of "0" to "127" that is supported but this data includes only values in the range of "60" to "115" When the offset value of -60 is applied to the data D1, there is generated transformed data having a value in the range of &quot; 0 &quot; to &quot; 55 &quot; By doing this, the degree of data compression to achieve can be realized. Accordingly, the generalized direct ODelta operator is provided with the data value of the full range in the data D1, that is, represented by 7 bits and conventionally represented by 8-bit bytes.

In accordance with the present invention, the direct ODelta value, i. E. Method 1, can be calculated for data with a positive value only using the procedure as outlined below except for the exemplary software code (lowValue = MIN = 0 and highValue = MAX = 127, wrapValue = 127 - 0 + 1 = 128).

Now, an example will be provided to further illustrate the above-mentioned ODelta operator. The original sequence of values is given in Equation 11 (Eq. 11) as follows.

식 11

Equation 11

The corresponding delta coding value is provided in Equation 12 (Eq. 12) as follows.

식 12

Equation 12

Corresponding direct ODelta coding values are provided in Equation 13 (Eq. 13) as follows.

식 13

Equation 13

Here, the wrapValue in the parameter is used.

The inverse ODelta operator, i.e., Method 1, can be used to generate an inverse ODelta value, e.g., as embodied by the exemplary software code below.

When this software code is executed and applied to Equation 13 (Eq. 13), this software generates values as provided in Equation 14 (Eq. 14).

식 14

Equation 14

This example uses wrapValue as the power of 2. This is not mandatory and wrapValue may be any value larger than the largest data value or wider than the used range if negative values are also available or if the range is changed by a pre-offset in a given sequence of data Value. Thereafter, there are additional examples showing this feature.

To summarize the foregoing with reference to FIG. 1, the present disclosure relates to encoder 10 and decoder 20. Optionally, encoder 10 and decoder 20 are combined and used as a codec, generally indicated by reference numeral 30. The encoder 10 may be operable to receive the original input data D1 that is encoded using the direct ODelta method, for example, to generate the corresponding encoded data D2 or D3. The encoded data D2 or D3 is optionally communicated over the communication network 40 or stored on a data storage medium 50 such as a data carrier such as an optical disk read only memory (ROM) or the like . The decoder 20 may be operable to receive encoded data D2 or D3, e.g., streamed over a communication network 40 or provided on a data storage medium 50, It may be operated to apply the inverse method, e.g. the inverse ODelta method, to produce substantially similar corresponding decoded data D5. Encoder 10 and decoder 20 may advantageously be implemented using digital hardware, such as, for example, a computing hardware capable of being operated to execute one or more software products, such as a codec provided as an exemplary embodiment in this description. . Alternatively, the encoder 10 and / or the decoder 20 are implemented using dedicated digital hardware.

As implemented in the encoder 10, the ODelta method utilizes the steps as shown in FIG. In an optional first step 100, the input data D1 is processed to find a range of values of data elements. In an optional second step 110, from the range of values, an offset, i. E., A pre-offset, is calculated to transform the data elements into a positive regime to produce the corresponding set of deformed elements. In a third step 120, the optionally modified elements in the second step 110 are then Direct ODelta encoded to produce corresponding ODelta encoded values. In a fourth step 130 the ODelta encoded values and optionally an offset value, a low value and / or a maximum value (highValue) are used to generate the data D3 from the data D2, Are then individually encoded using length-encoded (RLE), range-coded, or Hoffman-coded. Since the offset value, the low value (lowValue), and / or the maximum value (HighValue) can not always be compressed, a value is transmitted using the appropriate amount of bits from the encoder 10 to the decoder 20. Also, the offset value, the low value (lowValue), and / or the maximum value (highValue) are optional features for the direct ODelta operator; For example, the offset value may have a value of &quot; 0 &quot; in some situations, lowValue has a value MIN, and highValue has a value MAX, i.e. no deformation is applied and full range is not used. In particular, when a direct ODelta operator is implemented for 1-bit data, i. E. For bit-by-bit encoding, this operator does not need an offset value at all and then steps 100 and 110 are always ignored do. When an offset value is also used in step 110, the range value indicating the maximum and minimum values should be updated within step 110. [ The number of difference values, i. E., WrapValue, must also be known by the decoder 20, or else the encoder 10 must pass the value to the decoder 20 in the compressed data. Optionally, the default wrapValue (= highValue - lowValue + 1) is used in the encoder and in the decoder. Optionally, at least one of the encoder 10 and the decoder 20 may be configured to compress the data D1 to optimally compress the data D1 to generate encoded data D2, In a recursive manner.

The inverse ODelta method as implemented in the decoder 20 uses the steps as shown in FIG. In a first step 200, the data (D2 / D3 or D4) is decoded against the data used in step 130 described above to generate decoded ODelta data, where the decoded ODelta data is converted to an ODelta-encoded value And optionally has an individual offset value. In a second step 210, the ODelta-encoded values are decoded to produce a sequence of data elements. In a third step 220, the sequence of data elements is modified using an optional pre-offset value to generate decoded data D5; In some circumstances, this variant is set to a value of &quot; 0 &quot; i.e. the variant is not applied efficiently. Also, when performing, for example, a one-bit encoding or bit-by-bit encoding, it is possible to implement a method without using an offset value, thereby ignoring step 220. In addition, the decoder 20 must know the wrapValue to be able to decode the received data elements in an appropriate manner.

By using an offset to achieve only a positive value, more efficient data compression in data D2 or D3 can be achieved. If all the data values are already positive, there is no need to add any offset value. Of course, the negative offset values are optionally used to reduce the available range, as shown in the following example, but this is not mandatory.

The methods of FIGS. 2 and 3 can also be optionally optimized as well, by using only the available values that are ODelta coded. This optimization requires that the values used be known. For example, in the example of the above description, only values from 1 (= original minimum) to 126 (= original maximum) are provided in the original data set D1. The offset value is then 1 (-> lowValue = original minimum - offset = 1 - 1 = 0 and highValue = original maximum - offset = 126 - 1 = 125). When the pre-offset value is reduced from the original data D1, the following values are thereby generated in Eq. 15 (Eq. 15).

식 15

Equation 15

From Equation 15 (Eq. 15), the maximum value 125 can be either 125 or the wrapValue is the minimum value 126 (= number + 1 = highValue - lowValue + 1) (HighValue = original max - offset = 126 - 1 = 125). Now, it is necessary to store and / or transfer these values, and then the previous example can be modified by changing the process values as follows.

Corresponding direct ODelta operator values are provided in Equation 16 (Eq. 16).

식 16

Equation 16

It will be appreciated that all &quot; negative delta values &quot; are now decremented by a factor of 2 (i.e., = range change = 128 - 126). Similarly, in the decoder 20, the process values should be changed as follows.

Corresponding inverse ODelta values are provided in Equation 17 (Eq. 17) as follows.

식 17

Equation 17

When the pre-offset value is added to Equation 17 (Eq. 17), the following result in Equation 18 (Eq. 18) is obtained corresponding to the original data in Equation 15 (Eq. 15) as follows.

식 18

Equation 18

In this example, there is a relatively normal benefit derived by applying the direct ODelta operator to the offset and the maximum value (highValue) since the range of values is nearly filled. However, the reduction in entropy (E) can still be achieved, i.e., making the values in the frequency table or in the code table less when the values are properly propagated. The greatest gain can be achieved when the range is used less.

An exemplary embodiment of an actual 1-bit direct and inverse ODelta method, namely Method 1 or Method 3 for encoding and decoding data, is now provided by executable computer software code; This method uses the direct and inverse ODelta operators described above, Method 1 or Method 3. The software code, when executed on the computing hardware, can be operated to process bits from one byte buffer to another. In the software code, the GetBit, SetBit and ClearBit functions always update the HeaderBits value. The HeaderIndex value is also updated when the next bit is in the next byte. Optionally, the software code may be optimized to use only one set of HeaderIndex and HeaderBits values for the source and destination, so that the values are updated only when a given bit is written to the destination buffer.

The direct and inverse ODelta operators described above, Method 1 or Method 3, may be applied to any type of digital data, such as video data, video data, audio data, graphics data, seismic data, medical data, measurements, And is advantageously used to compress data. In addition, the one or more analog signals may also be compressed using the direct ODelta operator, for example, when converted to corresponding digital data for the first time, e.g., by using an ADC prior to compression. If the data is to be converted back to one or more analog signals, the DAC can be used after the operation when the inverse ODelta operator is used. However, although the direct ODelta operator alone is not generally effective for compressing data, it may be used in combination with other encoding methods such as variable length coding (VLC), arithmetic coding, range coding, run-length encoding, SRLE, entropy modifiers, It can provide effective data compression when used. This encoding method is used for the data D2 after the direct ODelta operator is used in the encoder 10. [ The encoded data D2 must be decoded again correspondingly before the resulting data is passed to the inverse ODelta operator implemented in the decoder 20. [ The ODelta operator can also be used for other types of entropy modifiers. In some situations, the direct ODelta operator increases the entropy (E), and the data compression algorithm can be used only when it provides useful data compression performance, for example, when it is selectively used based on the nature of the data being compressed, May be advantageously operated to selectively use the direct ODelta operator for use in encoding the data when selectively applied to selected portions of the input data D1 as described above.

A direct ODelta operator is devised for use in conjunction with a block encoder as disclosed in, for example, U.S. Patent Application No. 13 / 584,005, which is incorporated herein by reference, and the inverse ODelta operator is described in U.S. Patent Application No. 13 / 584,047, which is incorporated herein by reference in its entirety. Optionally, the direct ODelta operator and the inverse ODelta operator are advantageously used in combination with the multilevel coding method as disclosed in U.S. Patent Application No. 13 / 657,382, which is incorporated herein by reference. Advantageously, all types of 1-bit data, including binary states, for example in data D1, are processed by a 1-bit version of the Direct ODelta operator to produce the corresponding transformed data, Is then actually entropy encoded to produce the encoded data D2 or D3. Optionally, as described above, the direct ODelta operator is selectively used according to the nature of the original data D1.

Optionally, it can be realized using other methods of changing the entropy of the data before or after the direct ODelta operator. For example, the direct ODelta operator can also be used directly for multi-bit data in the generalized version of the direct ODelta operator. In addition, the 1-bit version of the direct ODelta operator described above is advantageously used for multi-bit data after all used bits are first provided in the serial sequence of bits.

When a plurality of methods are used for data compression in connection with the direct ODelta operator in the encoder 10, corresponding inverse operations are performed in reverse order, e.g., at the decoder 20.

The sequence of the following method is used in the encoder 10.

식 19

Equation 19

The inverse sequence of the following method is used in the decoder 20.

식 20

Expression 20

Where &quot; VLC &quot; represents variable length coding and &quot; EM &quot; represents entropy change.

The ODelta operator as described above is reversible and has no loss. In addition, the ODelta operator may perform, for example, bit-by-bit encoding, but may also be optionally implemented specifically for a one-bit data stream when performing for other data as well. Advantageously, all types of data can be processed using a generalized version of the direct ODelta operator. Advantageously, the direct ODelta operator is used when the data is compressed, and the corresponding inverse ODelta operator is used when the compressed data is decompressed. Optionally, when the ODelta operator is used, the direct ODelta operator and its corresponding inverse operation are used in reverse order, in other words, the inverse ODelta operator is first temporarily performed on the original bitstream, and then the original bit A direct ODelta operator is followed to generate the stream. One ODelta operator increases entropy, and the other ODelta operator reduces entropy. It is very rare that the direct ODelta operator should not change the entropy at all and then the inverse ODelta operator does not change the entropy. It should be noted that when the direct and inverse ODelta operators are used, for example, for Method 1, the inverse order of these operations is similar to the normal order of Method 4. A similar order change is also possible by Method 2 and Method 3.

In the 1-bit version, that is to encode the data in a bit-by-bit manner, the direct ODelta operator is advantageously disclosed without assuming a prediction, i. In the generalized version, the ODelta operator begins with a prediction that represents half of the available data range; For example, if the 5-bit is used for 32 different values of the input data value in the data D1, that is, the range of "0" to "31", the predicted value is 32/2 = 16. Advantageously, the ODelta operator needs to be provided with information about the available data range for the data elements to be processed using the operator.

The embodiments of the invention described in the foregoing description can reduce the entropy E provided to the data D1 as a bit or any digital value. The direct ODelta operator almost always provides improved entropy reduction compared to delta coding. If delta coding is used with byte wraparound and the difference ODelta operation for the original prediction (method 1) uses the values wrapValue = 256, lowValue = MIN = 0, and highValue = MAX = 255, Generate output results. If another direct ODelta method is used, or if the entire data range is not available in the input data, the ODelta operator produces better results by sending the selected method or lowValue and / or highValue, that is, . The smaller entropy allows the data to be compressed with a higher data compression ratio. The higher data compression ratio enables a smaller capacity data store to be used for a corresponding decrease in energy consumption and also allows for the use of slower data bandwidth when communicating compressed data.

In the foregoing, it will be appreciated that the type of difference and sum calculation is performed in the encoder 10 and the corresponding inverse calculation is performed in the decoder 20. [ It is also possible to use other prediction methods used in the encoder 10, and then a corresponding de-prediction is performed in the decoder 20. [ This actually means that there are at least four different direct ODelta methods as well as at least four corresponding inverse ODelta methods. A detailed and precise description of this method follows. Optionally, the calculation is performed in a recursive manner to obtain a higher-grade data compression in the encoded data D2 [or D3]. When performing this recursive computation, the range of numbers to be changed is used as a function of how many recursive computations are used. For example, in the encoder 10, the following calculation sequence is performed on the data D1 to generate the encoded data D2 [or D3].

식 21

Equation 21

And the corresponding inverse operations are performed in the decoder 20.

식 22

Equation 22

Each time data is represented by Equation 21 (Equation 21 corresponding to Method 1), Equation 22 (Equation 22 corresponding to Method 2), Equation 23 (Equation 23 corresponding to Method 3), and Equation 24 As indicated by Eq. 24), which corresponds to the number of times the data is processed, and since one of these methods can reduce the entropy of the data being processed much more than the other methods, It is optionally possible to try. When optimizing the use of the methods in the encoder 10 and / or the decoder 20, it is preferred to use the same or different methods a number of times when compared with the amount of information in the requested data, , It is advantageous to reduce entropy. Thus, methods 1 through 4 can be used to encode numerical values, where a "numerical value" is defined as the number of bits, as well as the multi-bit values, as in an encoded stream of bits by bit- Bit data as well as non-binary numbers within the &lt; / RTI &gt;

The difference operation represents the remainder of consecutive numeric values; Accordingly, the summation operation represents the sum of consecutive numerical values. These operations performed in the encoder 10 have their corresponding inverse operations at the decoder 20. The difference or sum may be calculated based on the current input value and the previous input or result value used as the predicted value. Other prediction values may also be used, and these values may use input and output values initially in the encoder to produce, for example, predictions, as long as they can be inverted to do so in the decoder as well.

Although none of these methods significantly compresses data in encoder 10 and decoder 20, all methods are advantageously used to reduce entropy, so that some other compression method may then be used to reduce entropy-reduced The data can be compressed more effectively. These other compression methods are optionally at least one of Hoffman coding, arithmetic coding, range coding, RLE coding, SRLE coding, and entropy modifier coding. However, for all methods, for example, if lossless compression of data and subsequent lossless decompression are achieved, then the operation and its inverse operation need to communicate small numeric values that can always be executed correctly. Of course, the encoder 10 and the decoder 20 have information as to what kind of numerical values are included in the input data D1. Advantageously, it is assumed that the numerical range defined by MIN and MAX is known. In principle, methods can always function directly based on existing data ranges. Numeric values that require operations are the lowest occurrence number value (lowValue) and the highest occurrence number value (highValue); lowValue is greater than or equal to MIN, and highValue is less than or equal to MAX.

Based on these values, other necessary numerical values may be derived. Advantageously, these values are communicated in various forms, wherein the lost values are advantageously computed. For example, if two values from the set ["lowValue", "highValue", "number"] are known, the "number" becomes [highValue - lowValue], and then the third value is calculated . Eliminating any values in the data D2 and then deriving these values in the decoder 20 can provide greater data compression in the data D2.

In addition to these values, a number P or &quot; prediction &quot; is required that can be used as the previous value in the calculation of the first value. The value between the &quot; 0 &quot; and the &quot; number &quot; value can always be selected for the number P, i.e. &quot; prediction &quot;. In addition, the above-described operations may be performed using a &quot; wrapValue &quot; algorithm to recoverably function when decoding data (D2 / D3 or D4) at the decoder 20, Value needs to be provided. However, this &quot; wrapValue &quot; should be larger than a &quot; number &quot;, advantageously, it should have a &quot; number &quot; value of +1. Optionally, depending on the nature of the data D1, the first &quot; prediction &quot; value may be selected, for example, to be &quot; 0 &quot; as described above if the data D1 is assumed to contain a value smaller than the larger value. . Alternatively, the first &quot; prediction &quot; value may be chosen to be equal to &quot; number &quot; if the data D1 is assumed to contain larger values than smaller values. If no assumption is made about the magnitude of the values, then it is preferable to use the value &quot; (wrapValue + 1) / 2 + lowValue &quot;

Now, an example of operations performed in the computing hardware when implementing the embodiments of the present disclosure will be described.

In the encoder 10, a first direct difference operation, i.e., method 1, is performed for all the data values, corresponding to the output value, i.e., &quot; result &quot;, input value, / RTI &gt;

Finally, the predicted value for the next input is set equal to the current input.

In the decoder 20, a first inverse difference operation, that is, method 1, is performed for all data values, corresponding to an output value, i.e., a result, an input value, i.e., an &quot; original & / RTI &gt;

Finally, the predicted value for the next input is set equal to the current result.

In the encoder 10, a second direct difference operation, that is, method 2, is performed for all the data values by using an output value, i.e., a result, calculated in a software loop, corresponding to the input value, / RTI &gt;

Finally, the predicted value for the next input is set equal to the current result.

In the decoder 20, the second inverse difference operation, that is, method 2, is performed for all the data values, corresponding to the output value, that is, the "result", the input value, that is, Quot; value. &Lt; / RTI &gt;

Finally, the predicted value for the next input is set to be equal to the current input as shown below.

In the encoder 10, a first direct sum operation, that is, method 3, is performed for all the data values, corresponding to input values, i.e., &quot; result &quot;, input values, Quot; value. &Lt; / RTI &gt;

Finally, the predicted value for the next input is set to be equal to the current input as shown below.

In the decoder 20, a first inverse sum operation, that is, method 3, is performed for all the data values by using an input value, i.e., a result, which corresponds to the input value, i.e., &quot;Quot; value. &Lt; / RTI &gt;

Finally, the predicted value for the next input is set to be equal to the current result as follows:

In the encoder 10, a second direct sum operation, that is, method 4, is performed for all the data values, corresponding to the input value, i.e., &quot; result &quot;, input value, Quot; value. &Lt; / RTI &gt;

Finally, the predicted value for the next input is set to be equal to the current result as follows:

In the decoder 20, the second inverse sum operation, that is, method 4, is performed on input values, i.e., &quot; result &quot;, corresponding to the input values, i.e., &quot;Quot; value. &Lt; / RTI &gt;

Finally, the predicted value for the next input is set to be equal to the current input as shown below.

This sum and difference operation, all four methods, i.e. when implementing the ODelta version of the encoder 10 and decoder 20, can also be applied to 1-bit data, i.e. bit-by-bit. In the context of 1-bit data, the following values are already known by both encoder 10 and decoder 20, i.e. MIN = 0, MAX = 1. Also, lowValue = MIN = 0, and highValue = MAX = 1 are beneficially assumed. Also in this case the &quot; number &quot; is therefore advantageously selected to be [highValue - lowValue = 1 - 0 = 1] and the wrapValue to be a &quot; number &quot; + 1 = 1 + 1 = Advantageously, the prediction value is chosen to be a &quot; 0 &quot; value, since there is only one bit of data under consideration, which may only have a positive value starting from lowValue = MIN = 0. For 1-bit data, Method 1 and Method 3 yield similar coding results to each other. Similarly, Method 2 and Method 4 yield similar coding results to each other. Having this knowledge advantageously simplifies the information that needs to be sent to the data D2, since various defaults can be assumed, and can be used for either method 1 or method 2 and for the selected prediction [input value (method 1) (D2, D3, or D4) to generate the decoded data D5 because the decoder 20 needs to send information on the number of times of the difference operation, ), It is possible to execute the correct inverse difference operation as many times as necessary.

The first example generated by using Method 1 or Method 3 to generate a similar output can also be processed by using either Method 2 or Method 4 to generate a similar output. The results shown below show that the data Eq. Lt; RTI ID = 0.0 &gt; 1, &lt; / RTI &gt;

At this time, the processed data has 24 &quot; 1 &quot; and 13 &quot; 0 &quot;, i.e. entropy is the same as in the first example, but the count of &quot; 1 &quot; This also does not always occur, unlike the often-occurring entropy changes between these different methods. For example, after four first elements of data, Method 1 and / or Method 3 generates three "1" s and one "0", while processing by original data and Method 2 and / Data has two &quot; 1 &quot; s and two &quot; 0 &quot; s. Thus, method 1 and / or method 3 may be smaller than method 2 and / or method 4 in this case, and may also produce smaller entropy than the original.

In a multi-bit implementation, if the data D1 includes a value in the range of -64 to +63, MIN = -64 and MAX = 63. By assuming that lowValue = MIN and highValue = MAX, "number" = 127 and wrapValue is advantageously chosen to be 128. However, when the data D1 changes randomly, the &quot; prediction &quot; value is advantageously set to the value [(wrapValue + 1) / 2 + lowValue = 64 + -64 = 0].

If the first value is, for example, -1, then the first value coded by the direct ODelta method and / or method 2 becomes -1 - 0 = -1 and accordingly, the direct ODelta method 3 and / It should be noted that the value is -1 + 0 = -1. For example, if the second value is 5, the direct ODelta method 1 produces 5 - 1 = 6 and the direct ODelta method 2 produces 5 - 1 = 6 Direct ODelta method 3 produces 5 + -1 = 4, and direct ODelta method 4 produces 5 + -1 = 4. Decoder 20 may in this case generate -1 + 0 = -1 as the first value when using the inverse ODelta method 1 and / or method 2 and use the inverse ODelta method 3 and / When used, -1 - 0 = -1 can be generated as the first value. Accordingly, the second value using the inverse ODelta method 1 becomes 6 + -1 = 5, and the second value becomes 6 + -1 = 5 by the inverse ODelta method 2. By the inverse ODelta method 3, 2 is 4 - -1 = 5, and the second value is 4 - -1 = 5 by the inverse ODelta method 4.

The solution can then be optimized if the numerical range actually contains only values between -20 and +27. In this exemplary case, it can be realized to transmit, for example, lowValue = -20 and highValue = 27. If both are sent, it can be realized that the number = 47, and then the wrapValue is advantageously chosen to be 48. Now, it can be realized to calculate the value [48/2 + -20 = 4] for prediction. The previous example then computes [-1 - 4 = -5] for the value -1 when, for example, ODelta method 1 or method 2 is used and [-1 + 4 = 3 ]. Similarly, the second value is (5 - 1) = 6, (5 - 5) = 10, (5 + -1) = 4, and (5 + 3) = 8 for the ODelta method. Decoding 20 again functions correctly and calculates a first value as -5 + 4 = -1 for Method 1 and / or Method 2, and calculates 3 + 4 = -1 for Method 3 and / And calculates a first value. Accordingly, the second value for the different method is decoded as (6 + -1) = 5, (10 + -5) = 5, (4 - 1) = 5, and (8 - 3) = 5.

All values in these examples are in the range, i.e., -64 to +63 or -20 to +27, and accordingly it is not necessary to perform the correction term in this exemplary value, but if any negative or positive change is sufficient If larger, a correction to the data value should be made by the given equations 21 to 24 (Eq. 21 to Eq. 24) in order to keep the result within the range. It should be noted here that the correction term refers to the wrap around value.

When lowValue is known, the coded values are useful to start with zero and end with a &quot; number &quot; value to simplify the coding table from which the entropy encoded data D3 from the encoder 10 to the decoder 20 should be sent Respectively. Such an operation is referred to as post-offset, and this post-offset must be removed from the coded data value before the inverse ODelta operation on data D4 after entropy decoding.

It is also possible to implement an offset with a pre-offset function, as described above, wherein the original input data D1 may already contain a value from 0 to &quot; number &quot; prior to actual execution of the ODelta method Is converted to a positive element. Also in this situation it is beneficial to carry out the information transmission in which this operation is required, in such a way that &quot; pre-offset &quot; and ODelta methods do not repeatedly transmit the same information or ignore what is already known due to some other method . This pre-offset result should be removed from the decoded data after the inverse ODelta operation to produce the appropriate D5 output data.

The above-described embodiments of the present invention can be modified without departing from the scope of the present invention as defined in the appended claims. The expressions "comprising," " including, "" integrating," "comprising," "having," It is to be understood that any item, component or element not expressly recited may also be present. Citation in singular form should also be construed as related to plural. The numbers included in parentheses in the appended claims are intended to be in the interest of understanding the scope of the claims and should not be construed as limiting the subject matter claimed in any way.

Claims (47)

An encoder (10) for encoding input data (D1) comprising a sequence of numerical values to produce corresponding encoded output data (D2 or D3)
A data processing device for applying to the input data (D1) a form of encoding of a difference, sum or both to produce one or more corresponding encoded sequences,
The one or more corresponding encoded sequences may be wrapped around with a maximum value wraparound processed or both wraparound processed to produce the encoded output data D2 or D3, (10). &Lt; / RTI &gt; 2. The apparatus of claim 1, wherein the data processing apparatus further comprises: one or more offset values for applying to the one or more corresponding encoded sequences for use in generating the encoded output data (D2 or D3) And to analyze the input data (D1), the one or more corresponding encoded sequences, or both, for computation. 3. The encoder (10) of claim 2, wherein the at least one offset value has a value of &quot; 0 &quot;. The encoder (10) of claim 1, wherein the encoder (10) is operable to process a numerical value comprising one or more 1-bit values, the encoder (10) Gt; (10) &lt; / RTI &gt; 2. The encoder (10) according to claim 1, wherein the one or more corresponding encoded sequences represent a change in sequential values of the input data (D1). 2. The encoder (10) according to claim 1, wherein the encoder (10) is operable to subdivide the input data (D1) into a plurality of data sections that are individually encoded. 7. A method according to claim 6, characterized in that the encoder (10) is operable to selectively apply an encoding to the data section only when data compression can be achieved in the encoded output data (D2 or D3) Lt; / RTI &gt; 2. The apparatus of claim 1, wherein the encoder (10) is operable to use a first default prediction value for a series of prediction values used to generate the encoded output data (D2 or D3) Lt; / RTI &gt; The method of claim 8, wherein the first default prediction value is at least one of "0", a minimum value (= lowValue), a maximum value (= highValue), (maximum value (highValue) + minimum value &Lt; / RTI &gt; 7. The apparatus of claim 6, wherein the encoder (10) is adapted to perform run-length of efficient mutually similar bits for encoding using run-length coding (RLE), Huffman coding, variable length encoding (VLE), range coding or arithmetic coding. And is thus operable to subdivide the input data (D1) into a plurality of sections. 11. The apparatus of any one of claims 1 to 10, wherein the processing device is implemented using computing hardware operable to execute one or more computer programs recorded on a machine-readable data storage medium Lt; / RTI &gt; A method of using an encoder (10) for encoding input data (D1) comprising a sequence of numerical values to generate corresponding encoded output data (D2 or D3)
(a) using a data processing apparatus to apply a difference, sum, or both type of encoding to the input data (D1) to generate one or more corresponding encoded sequences;
(b) in order to generate the encoded output data (D2 or D3), the one or more corresponding encoded sequences are wraparound processed to a maximum value, wrapped to a minimum value, or both wraparound processed (10). &Lt; RTI ID = 0.0 &gt; 11. &lt; / RTI &gt; 13. The method of claim 12, wherein the method further comprises: calculating one or more offset values for applying to the one or more corresponding encoded sequences for use in generating the encoded output data (D2 or D3) Using the data processing apparatus to analyze the input data (D1), the one or more corresponding encoded sequences, or both. 14. The method of claim 13, wherein the method comprises processing a numerical value comprising one or more 1-bit values, wherein the encoder (10) encodes the input data (D1) in a bit- Gt; (10) &lt; / RTI &gt; 13. The method of claim 12, wherein the method comprises using a data processing device to subdivide the input data (D1) into a plurality of data sections that are individually encoded. 16. The encoder of claim 15, wherein the method comprises selectively applying a form of encoding to the data section only when data compression can be achieved in the encoded output data (D2 or D3) (10). 13. The encoder (10) of claim 12, wherein the method comprises using a first default prediction value for a series of prediction values used to generate the encoded output data (D2 or D3) ). 18. The method of any one of claims 12 to 17, wherein the method further comprises implementing the data processing apparatus using computing hardware operable to execute one or more computer programs recorded on a machine-readable data storage medium &Lt; / RTI &gt; further comprising the step of: A decoder (20) for decoding encoded data (D2 or D3) to produce corresponding decoded output data (D5)
The decoder (20)
And a data processing device for processing one or more portions of the encoded data (D2 or D3)
The data processing apparatus may be operable to apply a difference, sum, or both type of decoding to one or more corresponding encoded sequences of the one or more portions, and to generate decoded output data (D5) The above encoded sequence is characterized in that the maximum value is wraparound processed, the minimum value is wraparound processed, or both are wraparound processed. 20. The apparatus of claim 19, wherein the decoder (20) is operable to decode encoded data (D2 or D3) comprising one or more 1-bit values and the decoder (20) And is operable to decode the encoded data (D2 or D3). 20. The apparatus of claim 19, wherein the data processing apparatus is operative to receive one or more offset values, a minimum or maximum value, for applying to the one or more encoded sequences for use in generating the decoded output data (D5) Gt; (20) &lt; / RTI &gt; 23. The decoder (20) of claim 21, wherein the one or more offset values have a value of &quot; 0 &quot;. 20. The decoder (20) according to claim 19, wherein the one or more corresponding encoded sequences represent a change in a sequential value encoded with the encoded data (D2 or D3). 20. The decoder (20) of claim 19, wherein the data processing device is operable to estimate a default value of a first predictor value in a series of data to be decoded. 25. The decoder (20) of claim 24, wherein the default value has a &quot; 0 &quot;. 26. A data processing apparatus according to any one of claims 19 to 25, characterized in that it is implemented using computing hardware operable to execute one or more computer programs recorded on a machine-readable data storage medium (20). A method of using a decoder (20) for decoding encoded data (D2 or D3) to produce corresponding decoded output data (D5)
The method comprises:
Using a data processing device to process one or more portions of the encoded data (D2 or D3)
The data processing apparatus may be operable to apply a difference, sum, or both type of decoding to one or more corresponding encoded sequences of the one or more portions, and to generate decoded output data (D5) Wherein the encoded sequence is characterized in that the maximum value is wraparound processed, the minimum value is wraparound processed, or both are wraparound processed. 28. The method of claim 27, wherein the method comprises operating a decoder (20) for decoding encoded data (D2 or D3) comprising one or more 1-bit values, the decoder (20) (D2 or D3) in a bit-by-bit manner. &Lt; Desc / Clms Page number 12 &gt; 28. The apparatus of claim 27, wherein the method further comprises: providing the data processing apparatus to calculate one or more offset values, a minimum and a maximum, for applying to the one or more encoded sequences for use in generating the decoded output data (D5) The method comprising the steps of: operating the decoder (20). 28. The method of claim 27, wherein the method comprises operating the data processing apparatus to estimate a default value of a first predictor value in a series of data to be decoded. 32. The method according to any one of claims 27 to 30,
The method includes implementing the data processing apparatus using computing hardware operable to execute one or more computer programs recorded on a machine-readable data storage medium Way. In the codec 30,
At least one encoder (10) according to claim 1 encoding the input data (D1) to generate corresponding encoded data (D2)
A codec (30) comprising at least one decoder (20) according to claim 19 for decoding encoded data (D2 or D3) to produce corresponding decoded data (D5). A computer-readable recording medium containing a computer program,
Wherein the computer program is executable on computing hardware to perform the data encoding method of any one of claims 12 to 17. A computer-readable recording medium containing a computer program,
Wherein the computer program is executable on computing hardware for carrying out the data decoding method of any one of claims 27 to 30. delete delete delete delete delete delete delete delete delete delete delete delete delete

Images